TMPGEnc - same video, different processors = different final compressed files sizes

[Hulk]

I just noticed that TMPGEnc encodes differently on different CPU's.

I just encoded a hour and a half long video on my P4 1.9 desktop and my laptop that has a PIII 1.2.

Templates were exactly the same for both. CQ80, 620x480, etc...

Anyway the P4 file ended up being about 3.1GB while the PIII one was about 3.3GB. I guess different routines are used depending on the processor detected.

For the record, the PIII encoded at 65% the rate of the P4. Exactly what you would expect given both cpu frequencies.

Edit - Sorry about that, I'm getting old!

[dchip]

Re: TMPGEnc - same video, different processors = different final compressed files sizes

Originally posted by Hulk
[B]I Anyway the P4 file ended up being about 3.1GB while the PIII one was about 3.1GB.

These files sound identical.

[paulw]

No.. He said.

"Anyway the P4 file ended up being about 3.1GB while the PIII one was about 3.3GB. "

[dchip]

That is odd. Same OS on both systems? Also, keep in mind that reported filesizes in in utilities like Windows Explorer often differ from the actual filesize. I have made some DV AVI's that show as different sizes in Windows Explorer and the Properties for the files. Another possiblility is that TMPGEnc does do something different with a slower/faster processor.

paulw: It's all about what time I saw the post.

[Hulk]

Sorry about the confusion caused by my typo of not initially listing the file sizes correctly.

The desktop is running Windows XP Pro and the laptop (Compag EVO n600c) is running Windows 2000 Pro.

After compressing the same 22GB file using the same TMPGEnc template I brought the file compressed using the laptop over to the desktop. Believe me, I was looking at them side by side and the difference was about 200 MB, not a few bytes or bits.

In addition, since I have these computers networked they were both using the exact same source file.

Does TMPGEnc use SSE II? That could be the difference. I don't know it seems weird.

[Dr Mordrid]

Yes, TMPGEnc can use SSE2 if the CPU supports it. Otherwise it'll use SSE or even MMX.

Dr. Mordrid

[Hulk]

If the routines are different for SSE and SSEII that could explain it, I guess. Between a slightly different algorithm and round-off error.

[Brian Ellis]

Do you have the same (hell! I've forgotten the correct term ), you know, the size of the slice of disk that even 1 byte will take, causing slack?

[droopy]

@Brian,
That doesn't make a difference and is only of influence when using a lot of small files. This is one big 3.1 / 3.3 GB file?

@All
What different CPU's support is of no influence here I'd guess. This would only make a difference in the speed data is processed.

MPEG compression is lossy compression and partly a judgement call. I can therefore imagine that on difference occassions for the same data an algorythm could result in different outcomes. I know I see it already when editting in Photoshop and saving as JPEG. Same image, same quality, different file sizes / content, albeit small differences.

Regards, Leon

[thop]

i think Brian Ellis means block size. does one have FAT32 and the other NTFS?

[Hulk]

Both are NTFS but that wouldn't matter, even if the cluster sizes were differenct once I brought both files to the same drive.

This was a long video so I guess small compression differences could add up.